nyancrimew
/
LatinIME
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

am 758d0936: Get entry count after truncation using LanguageModelDictContent. 

* commit '758d09364457b9d3d0c514a7fcfc8a6e317c9222':
  Get entry count after truncation using LanguageModelDictContent.
main
Keisuke Kuroyanagi 2014-08-27 11:22:37 +00:00 committed by Android Git Automerger
parent 445f5536b6 758d093644
commit 276ce6c0fd
3 changed files with 26 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
native/jni/src/suggest/policyimpl/dictionary/structure/v4/content/language_model_dict_content.cpp
View File

@ -23,6 +23,9 @@
namespace latinime {
const int LanguageModelDictContent::UNIGRAM_COUNT_INDEX_IN_ENTRY_COUNT_TABLE = 0;
const int LanguageModelDictContent::BIGRAM_COUNT_INDEX_IN_ENTRY_COUNT_TABLE = 1;
bool LanguageModelDictContent::save(FILE *const file) const {
return mTrieMap.save(file);
}
@ -78,12 +81,15 @@ LanguageModelDictContent::EntryRange LanguageModelDictContent::getProbabilityEnt
}
bool LanguageModelDictContent::truncateEntries(const int *const entryCounts,
const int *const maxEntryCounts, const HeaderPolicy *const headerPolicy) {
const int *const maxEntryCounts, const HeaderPolicy *const headerPolicy,
int *const outEntryCounts) {
for (int i = 0; i <= MAX_PREV_WORD_COUNT_FOR_N_GRAM; ++i) {
if (entryCounts[i] <= maxEntryCounts[i]) {
outEntryCounts[i] = entryCounts[i];
continue;
}
if (!turncateEntriesInSpecifiedLevel(headerPolicy, maxEntryCounts[i], i)) {
if (!turncateEntriesInSpecifiedLevel(headerPolicy, maxEntryCounts[i], i,
&outEntryCounts[i])) {
return false;
}
}
@ -185,7 +191,8 @@ bool LanguageModelDictContent::updateAllProbabilityEntriesInner(const int bitmap
}
bool LanguageModelDictContent::turncateEntriesInSpecifiedLevel(
const HeaderPolicy *const headerPolicy, const int maxEntryCount, const int targetLevel) {
const HeaderPolicy *const headerPolicy, const int maxEntryCount, const int targetLevel,
int *const outEntryCount) {
std::vector<int> prevWordIds;
std::vector<EntryInfoToTurncate> entryInfoVector;
if (!getEntryInfo(headerPolicy, targetLevel, mTrieMap.getRootBitmapEntryIndex(),
@ -193,8 +200,10 @@ bool LanguageModelDictContent::turncateEntriesInSpecifiedLevel(
return false;
}
if (static_cast<int>(entryInfoVector.size()) <= maxEntryCount) {
*outEntryCount = static_cast<int>(entryInfoVector.size());
return true;
}
*outEntryCount = maxEntryCount;
const int entryCountToRemove = static_cast<int>(entryInfoVector.size()) - maxEntryCount;
std::partial_sort(entryInfoVector.begin(), entryInfoVector.begin() + entryCountToRemove,
entryInfoVector.end(),

7
native/jni/src/suggest/policyimpl/dictionary/structure/v4/content/language_model_dict_content.h
View File

@ -39,6 +39,9 @@ class HeaderPolicy;
*/
class LanguageModelDictContent {
public:
static const int UNIGRAM_COUNT_INDEX_IN_ENTRY_COUNT_TABLE;
static const int BIGRAM_COUNT_INDEX_IN_ENTRY_COUNT_TABLE;
// Pair of word id and probability entry used for iteration.
class WordIdAndProbabilityEntry {
public:
@ -158,7 +161,7 @@ class LanguageModelDictContent {
// entryCounts should be created by updateAllProbabilityEntries.
bool truncateEntries(const int *const entryCounts, const int *const maxEntryCounts,
const HeaderPolicy *const headerPolicy);
const HeaderPolicy *const headerPolicy, int *const outEntryCounts);
private:
DISALLOW_COPY_AND_ASSIGN(LanguageModelDictContent);
@ -197,7 +200,7 @@ class LanguageModelDictContent {
bool updateAllProbabilityEntriesInner(const int bitmapEntryIndex, const int level,
const HeaderPolicy *const headerPolicy, int *const outEntryCounts);
bool turncateEntriesInSpecifiedLevel(const HeaderPolicy *const headerPolicy,
const int maxEntryCount, const int targetLevel);
const int maxEntryCount, const int targetLevel, int *const outEntryCount);
bool getEntryInfo(const HeaderPolicy *const headerPolicy, const int targetLevel,
const int bitmapEntryIndex, std::vector<int> *const prevWordIds,
std::vector<EntryInfoToTurncate> *const outEntryInfo) const;

13
native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_writing_helper.cpp
View File

@ -93,14 +93,16 @@ bool Ver4PatriciaTrieWritingHelper::runGC(const int rootPtNodeArrayPos,
}
if (headerPolicy->isDecayingDict()) {
int maxEntryCountTable[MAX_PREV_WORD_COUNT_FOR_N_GRAM + 1];
maxEntryCountTable[0] = headerPolicy->getMaxUnigramCount();
maxEntryCountTable[1] = headerPolicy->getMaxBigramCount();
maxEntryCountTable[LanguageModelDictContent::UNIGRAM_COUNT_INDEX_IN_ENTRY_COUNT_TABLE] =
headerPolicy->getMaxUnigramCount();
maxEntryCountTable[LanguageModelDictContent::BIGRAM_COUNT_INDEX_IN_ENTRY_COUNT_TABLE] =
headerPolicy->getMaxBigramCount();
for (size_t i = 2; i < NELEMS(maxEntryCountTable); ++i) {
// TODO: Have max n-gram count.
maxEntryCountTable[i] = headerPolicy->getMaxBigramCount();
}
if (!mBuffers->getMutableLanguageModelDictContent()->truncateEntries(entryCountTable,
maxEntryCountTable, headerPolicy)) {
maxEntryCountTable, headerPolicy, entryCountTable)) {
AKLOGE("Failed to truncate entries in language model dict content.");
return false;
}
@ -204,7 +206,10 @@ bool Ver4PatriciaTrieWritingHelper::runGC(const int rootPtNodeArrayPos,
&traversePolicyToUpdateAllPtNodeFlagsAndTerminalIds)) {
return false;
}
*outUnigramCount = traversePolicyToUpdateAllPositionFields.getUnigramCount();
*outUnigramCount =
entryCountTable[LanguageModelDictContent::UNIGRAM_COUNT_INDEX_IN_ENTRY_COUNT_TABLE];
*outBigramCount =
entryCountTable[LanguageModelDictContent::BIGRAM_COUNT_INDEX_IN_ENTRY_COUNT_TABLE];
return true;
}